Layer-0 failure recovery is important in optical networking. Best in class optical networking deployments provide automatic failure recovery
With increasing demands for bandwidth in optical networks, optical networking technologies are evolving to transmit more bits per second (bps) over optical links. The optical spectrum employed has been standardized such as in ITU-T Recommendation G.694.1 (June 2002) “Spectral Grids for WDM Applications: DWDM Frequency Grid” and ITU-T Recommendation G.698.2 (November 2009) “Amplified Multichannel DWDM Applications with Single Channel Optical Interfaces,” the contents of which are incorporated by reference herein. The optical spectrum can be segmented into transmission windows at different wavelengths such as a “C band” which is from about 1530 to about 1565 nm and which corresponds to the gain bandwidth of Erbium Doped Fiber Amplifiers (EDFAs). Other transmission windows can include an “L band” (from about 1565 to about 1625 nm), an “S band” (from about 1460 to about 1530 nm), etc. Conventionally, Dense Wavelength Division Multiplexed (DWDM) networks typically use a fixed bandwidth (e.g., 25, 50, 100, or 200 GHz) centered on the ITU grid (i.e., center channel wavelength/frequency) for each channel. This can be referred to as a “gridded DWDM optical spectrum,” i.e., each channel occupies a spot on the grid in an associated transmission window. However, with higher number of bits per second being conveyed (especially beyond 100 Gbps), it is getting increasingly complex to fit channels within a fixed slotted spectral bandwidth. Allocating larger bandwidth channels is required for high baud rate signals, allocation which cannot respect the fixed grid ITU spectrum. Accordingly, there is a movement towards “gridless” or flexible DWDM spectrum where the (slot) width of the channels is flexible and/or variable (i.e., the slot width is uncertain before a slot is allocated), also known as flex grid. Optimal spectrum utilization or spectral mining is another reason to move towards flexible spectrum solutions where more channels can be grouped together without guard bands in between, groups which can potentially occupy the full DWDM spectral bandwidth. For comparisons, whereas in a conventional gridded system each channel has a predetermined bandwidth, i.e. 25, 50, 100, or 200 GHz, in flexible spectrum systems each channel has a variable bandwidth of N GHz, where N can be any amount of bandwidth and can be different for each channel.
FIG. 1 is a graph showing an example of a flexible spectrum 10 employed in an optical system. The flexible spectrum 10 illustrated includes four channels 12, 14, 16, 18. The first two channels 12, 14 occupy 50 GHz of bandwidth (wide) each with guard bands 20 (spaces) therebetween. For clarity, a conventional gridded system using 50 GHz spacing would include each channel on the optical spectrum being similar to channels 12, 14 shown. However, employing flexible spectrum, third channel 16 occupies 400 GHz (width) and can be, for example, a 2 Tbps signal. The fourth channel 18 is a 4×100 Gbps signal with each of the 100 Gbps signals occupying 37.5 GHz (width) for a total of 150 GHz (width). Channels 16, 18 can be referred to as ‘super’ channels and will be more common as more advanced modulation techniques are utilized to increase the number of bits per second conveyed.
In conventional gridded systems, adding, or deleting a single channel has minimal impact on other existing in-service channels particularly when there are many such channels in such gridded systems; the guard spaces help. However, capacity change (i.e., adding or deleting a channel) has significantly more impact in flexible spectrum networks since such capacity change is no longer adding one channel among many as in gridded systems, but could be adding or deleting a significant portion of the spectrum. For example, adding or deleting either channel 16, 18 (with limited guard space) will have significant impact on in-service channels 12, 14.
However, it is highly desirable, in either fixed grid or flexible grid optical systems, particularly while adding new bandwidth capacity along an optical path for lighting new photonic services to provide the additional bandwidth capacity without impairing the performance of existing photonic services already on that optical path. Conversely, while deleting bandwidth capacity along an optical link, in either fixed grid or flexible grid optical systems, extinguishing photonic services cannot impair the performance of remaining provisioned services on that optical link.
The time taken to complete restoration of a photonic service to a protection path is constrained partly due to optical hardware limitations to achieve a stable light output, and mostly due to the requirement of making only non-service affecting capacity changes on non-faulted optical links providing the restoration path. For clarity, per channel actuators in optical hardware can light up additional wavelengths relatively fast (typically in milliseconds) to provision a photonic service, however doing so each newly lit channel can create spectral gain cross-talk in neighboring in-service channels.
When new optical signals (photonic services) are added to a path, due to changes in spectral loading over each optical link, the change triggers multiple physical effects which can result in a change in power and Optical Signal to Noise Ratio (OSNR) levels of pre-existing channels. Such physical effects imparted to existing channels are primarily caused by at least one of the following three factors:
Amplifier gain ripple and tilt: Different spectral loadings will result in different power gains at different frequencies over the spectrum;
Stimulated Raman Scattering (SRS) effect which appears especially with high per channel launch powers; and
Spectral Hole Burning (SHB) effect which creates a photonic energy starving hole in the spectrum at the output of EDFAs which becomes more prevalent in low spectral loaded scenarios since the holes tend to disappear under heavy pumping levels.
A cumulative effect from all the factors highlighted above contributes to physical power level perturbations (glitches) experienced by existing provisioned photonic services, which unchecked can either result in undershoots causing lower OSNR for existing photonic services or overshoots causing non-linear impairments to neighboring photonic services. The magnitude of such physical perturbations further depends on the number of amplifiers in the optical path across the optical network and the spectral placement of the existing and newly added channels along the optical path.
FIGS. 2A, 2B, 3A and 3B illustrate examples of conventional Spectral Hole Burning (SHB) effects affecting existing photonic services due to an addition change in capacity.
FIGS. 2A and 2B show an example of the SHB effect on the power level of a pre-existing optical channel provisioned in an optical link (along an optical path). The in-service channel is represented by the power level signal peak above (Amplified Spontaneous Emission) ASE light background in FIG. 2A with some negative tilt applied. The existing in-service channel is provisioned at a target power shown in dashed line. While a group of channels is added in an energy starved portion of the spectrum, the power level of the existing channel(s) can overshoot as shown in FIG. 2B. With the amplifier over-gained due to compensation at the spectral hole, the power in the spectral hole is over-represented. This results in the existing in-service channel being over-powered.
FIGS. 3A and 3B show another example of the SHB effect on the power level of a pre-existing optical channel provisioned in a link (along an optical path). The in-service channel is provisioned at target power (shown in dashed line) in a low ASE light portion of the spectrum in FIG. 3A with some negative tilt applied. While a group of channels is added in an energy starved portion of the spectrum, the power level of the existing channel(s) can undershoot as shown in FIG. 3B. With the amplifier over-gained due to compensation of the spectral hole, the power in the spectral hole is over-represented. This (and the negative tilt) results in the existing in-service channel being under-powered.
In general, depending on where an existing in-service channel is located in the optical spectrum and depending where a channel is added in the spectrum different impairments can be experienced by the existing in-service channel. For certainty, while not shown, deleting a channel can and often does have an impairment on other existing in-service channels. In this sense, while per channel actuators in optical hardware can switch on additional wavelengths fast (typically in milliseconds), doing so can create spectral gain cross-talk (power glitches) in neighboring in-service channels.
The activation of per-channel actuators is typically slowed down in a controlled way to switch on in seconds instead of milliseconds. Examples of such methods include U.S. Pat. No. 9,344,191 entitled “Systems and Methods for Capacity Changes in DWDM Networks Including Flexible Spectrum Systems” issued May 17, 2016, the entirety of which is incorporated herein by reference. In brief, when adding (or restoring) channels to an optical path, the controllers need to consider the potential physical impairments on existing channels by: adding small increments of power for adding channels in the desired part of the spectrum, slowing down to monitor the power impairments on existing channels if any, and nullifying those impairments before adding more power into the spectrum. U.S. Pat. No. 9,344,191 describes a photonic controller configured to switch and move per channel actuators from one attenuation level to another in seconds, controller which employs live feedback data from optical per channel power monitors adding additional overheads for adding or deleting capacities in a network.
Therefore, capacity changes in an optical transmission line system remain a challenge. In particular, there is a need to improve optical path restoration functionality.
Further, conventional approaches do not consider the fact that the margin impact by control speeds is not uniform across a path, i.e., the SNR margin impact on existing in-service channels on a path due to power offsets created by a faster control speed on that path may not be the same over the entire path if channel population or spectral loading condition varies along the path.